we provide a set of pre-defined gym environments for various tasks. The action spaces are various from discrete to continuous and the observation spaces are various from depth image to RGB-D image.
The details about these environments are shown in the register file. We summarize the environments as below:
Task: find target object and avoid obstacle simultaneously.
UnrealEnv: RealisticRendering
Naming rule: {Task}-{Scene}{Target}{ActionSpace}-{Version}
- Search-RrDoorDiscrete-v0
- Search-RrDoorContinuous-v0
- Search-RrPlantsDiscrete-v0
- Search-RrPlantsContinuous-v0
- Search-RrSocketsDiscrete-v0
- Search-RrSocketsContinuous-v0
Task: actively track the target object.
Naming rule: {Task}-{Scene}{Target}{PathID}{AugmentEnv}-{Versrion}
- Tracking-City1StefaniPath1Random-v0
- Tracking-City1StefaniPath1Static-v0
- Tracking-City1MalcomPath1Static-v0
- Tracking-City1StefaniPath2Static-v0
- Tracking-City2MalcomPath2Static-v0
Active Object Tracking Environment is used in
Luo, Wenhan, et al. "End-to-end Active Object Tracking via Reinforcement Learning." arXiv.
More details about the environment definition can be found in this paper.